PCNA Complex and Implications for Clamp Sliding During DNA Replication and Repair

ARTICLE

Received 6 Jun 2014 | Accepted 29 Jan 2015 | Published 12 Mar 2015 DOI: 10.1038/ncomms7439 Structure of p15PAF–PCNA complex and implications for clamp sliding during DNA replication and repair

Alfredo De Biasio1,w, Alain Iba´n˜ez de Opakua1, Gulnahar B. Mortuza2,3, Rafael Molina2, Tiago N. Cordeiro4, Francisco Castillo5, Maider Villate1, Nekane Merino1, Sandra Delgado1, David Gil-Carto´n1, Irene Luque5, Tammo Diercks1, Pau Bernado´4, Guillermo Montoya2,5 & Francisco J. Blanco1,6

The intrinsically disordered protein p15PAF regulates DNA replication and repair by binding to the proliferating cell nuclear antigen (PCNA) sliding clamp. We present the structure of the human p15PAF–PCNA complex. Crystallography and NMR show the central PCNA-interacting protein motif (PIP-box) of p15PAF tightly bound to the front-face of PCNA. In contrast to other PCNA-interacting proteins, p15PAF also contacts the inside of, and passes through, the PCNA ring. The disordered p15PAF termini emerge at opposite faces of the ring, but remain protected from 20S proteasomal degradation. Both free and PCNA-bound p15PAF binds DNA mainly through its histone-like N-terminal tail, while PCNA does not, and a model of the ternary complex with DNA inside the PCNA ring is consistent with electron micrographs. We propose that p15PAF acts as a ﬂexible drag that regulates PCNA sliding along the DNA and facilitates the switch from replicative to translesion synthesis polymerase binding.

1 Structural Biology Unit, CIC bioGUNE, Parque Tecnolo´gico de Bizkaia Edificio 800, 48160 Derio, Spain. 2 Structural Biology and Biocomputing Programme, Centro Nacional de Investigaciones Oncolo´gicas, Melchor Fernańdez Almagro 3, 28029 Madrid, Spain. 3 Protein Structure and Function Program, Novo Nordisk Foundation Center for Protein Research, Faculty of Health and Medical Sciences, University of Copenhagen, Blegdamsvej 3B, 2200 Copenhagen, Denmark. 4 Centre de Biochimie Structurale, INSERM U1054, CNRS UMR 5048, Universite´ Montpellier 1 and 2, 29 rue de Navacelles, 34090 Montpellier, France. 5 Department of Physical Chemistry and Institute of Biotechnology, Universidad de Granada, Fuentenueva s/n, 18071 Granada, Spain. 6 IKERBASQUE, Basque Foundation for Science, MarıáDıáz de Haro 3, 48013 Bilbao, Spain. w Present address: Elettra Sincrotrone Trieste, Trieste, Italy. Correspondence and requests for materials should be addressed to A.De.B (email: [email protected]) or to F.J.B. (email: [email protected]).

NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications 1 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439

he DNA-sliding clamp proliferating cell nuclear antigen binding to equivalent sites in the PCNA trimer, with a dissocia- (PCNA) is an essential factor in replication and repair that tion constant of 1.1 mMat25°C (Fig. 1c). Thus, three p15 Trecruits polymerases and other DNA-modifying enzymes molecules bind to the three PCNA protomers, with no indication to the replication fork1. PCNA forms an 86-kDa homotrimeric of cooperativity. The enthalpic contribution to the free energy ring2 that encircles the DNA and anchors binding partners, is large, but so is the entropic cost, in line with a loss in p15 preventing them from falling off the genomic template. The ring conformational freedom on binding to PCNA. has an inner diameter of B35 Å, much wider than the B24 Å As a disordered protein, p15 yields NMR spectra with sharp diameter of a DNA duplex in the B-form3. Apart from DNA- signals and poor dispersion in the proton frequency. On addition editing factors, PCNA also interacts with non-enzymatic proteins of excess PCNA, signal intensities decrease generally and most that regulate cell cycle control and survival4,5. Many PCNA- notably for the central residues V53-D75 that even disappear interacting proteins bind through the PIP-box, a consensus (Fig. 1d). This observation highlights the central region of p15 as sequence with the pattern QXXhXXaa, where h is an aliphatic- the principal binding site for PCNA, which includes the PIP-box hydrophobic residue, a aromatic-hydrophobic and X any motif (62QKGIGEFF69, conserved residues underlined). The p15 residue6. In all known co-crystal structures, the PIP-box NMR signals still visible in the complex retain their low residues bind to the PCNA front-face forming a short amino- dispersion, indicating that they remain largely disordered and terminal (N-terminal) b-strand that interacts with the PCNA flexible. Nevertheless, in the presence of PCNA, two segments in C-terminus, a short 310 helix placed in a hydrophobic pocket and the N- and C- terminal regions of p15 around the conserved basic a carboxyl-terminal (C-terminal) b-strand of variable length that residues R14K15 and P100, respectively, show significantly 7 15 pairs with the PCNA interdomain connecting loop (IDCL) . reduced signal intensities (Fig. 1d) and increased NR2 The recognition surface on PCNA is small, contacting 11–18 relaxation rates (Fig. 1e). While all R2 rates increase due to the residues of the ligand. The regions flanking these tightly bound larger molecular size, their abnormal increase in these two regions residues are usually not visible in electron density maps, indicates some conformational restrictions in the complex. Both suggesting that the chain breaks away from the clamp surface. regions contain clusters of positively charged amino acids that A common feature of the PIP-box region in these proteins is its might interact transiently and unspecifically with the negatively disordered nature, providing an adaptable, yet specific contact site charged PCNA surface. In line with this, chemical shift whose affinity for PCNA can be tuned by subtle changes in the perturbations (CSP) here are the largest observed outside the sequence8. core binding site (Fig. 1f), but still so small that only weak The PCNA-associated factor p15 (p15PAF, hereafter named transient interaction is expected. Changes in the NMR spectrum p15) is a nuclear protein with a PIP-box9 and is involved in the of PCNA on p15 binding are extensive (Fig. 1b), more than seen regulation of DNA replication and cell cycle progression by when bound to a PIP-box fragment of p21CIP1 (hereafter named interacting with PCNA10–12. Overexpression of p15 correlates p21)18, suggesting a larger binding area for p15. with tumour progression and poor prognosis in a number of To study the role of different p15 regions in the interaction human cancers13–16. Evidence for p15–PCNA interaction comes with PCNA, we designed three p15 fragments: the N-terminal from yeast two-hybrid and co-immunoprecipitation experiments, deletion mutant p1532–111 (hereafter named p15DN) lacking but no structural detail of the direct interaction is known. We many of the residues experiencing reduced NMR signal intensity 15 50–77 have previously shown that p15 is a monomeric intrinsically and enhanced NR2 rates on PCNA binding; the p15 disordered protein with conformational preferences in certain fragment comprising the central residues that experience the regions, including the PIP-box17. largest drop in NMR signal intensity on PCNA binding; and Here we present an extensive characterization of the interac- p1559–70 that corresponds to the smallest PIP-box containing tion between p15 and PCNA based on an integrative structural fragment of p21 with reported binding to PCNA19. approach. Three p15 molecules bind to the trimeric PCNA ring All three p15 fragments bind PCNA, but with decreasing with low micromolar affinity. The central region of p15, including affinity (Supplementary Fig. 1 and Supplementary Table 1). The the PIP-box, is tightly bound to the front-face and inner surface N-terminal deletion mutant binds PCNA with only slightly lower of PCNA, while the N- and C-termini remain predominantly affinity than p15, implying that any transient interactions of disordered at its back- and front-face, respectively, with transient the N-terminal V2-S31 residues stabilize the complex only interactions at the PCNA outer surface. This arrangement is marginally. The affinity of p1550–77, however, is reduced by a consistent with the protection from degradation by the 20S factor of 5 and the shortest core fragment p1559–70 even by a proteasome particle, which rapidly degrades free p15. Most factor of 33. These results indicate that residues G50-T58 and/or importantly, we show that p15 binds DNA, mainly through its L71-E77 strongly contribute to PCNA binding, which is further histone-like N-terminal tail. While PCNA alone does not interact enhanced by residues A32-G49. with DNA, p15 binds to PCNA and DNA simultaneously via The NMR spectra of PCNA in the presence of the different independent sites. Molecular modelling and electron microscopy constructs indicate rather similar binding modes for p15, p15DN show that a ternary complex with both p15 and DNA inside the and p1550–77, but less so for p1559–70 (Supplementary Fig. 2). The PCNA ring can form. We discuss the possible implications of strong intensity reduction in many PCNA signals in the presence these findings for the regulation of DNA replication and repair, of p15 (and to a lesser extent in the presence of p15DN) most and propose that p15 acts as a flexible drag that regulates PCNA likely derives from line broadening caused by the increased size of sliding along the DNA, facilitating the switch from replicative to the complex (125 kDa) and by transient interactions with the p15 translesion synthesis polymerase binding. chain termini. While we have previously assigned the NMR spectrum of PCNA free and bound to a p21 PIP-box peptide18,20, p15 excess leads to prohibitively large intensity reduction and line Results broadening that precluded assignment of the p15-bound PCNA p15–PCNA interaction involves an extended PIP-box region. spectrum by triple resonance methods. Instead, we transferred the Changes in the NMR spectra of p15 and PCNA in the presence of assignment from free PCNA, based on extrapolation from signal each other (Fig. 1a,b) confirm a direct interaction between both shifting in a titration with p15. In this way, we obtained many proteins. Isothermal titration calorimetry (ITC) measurements plausible quantitative CSP values for PCNA signals on p15 could be fitted very well to a model assuming independent p15 binding and assigned an arbitrary CSP cutoff value where signals

2 NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439 ARTICLE

ab d 20 105 G49 G50 15 110 110 G64 p15 115 10 G66 p15–PCNA

N (p.p.m.) 120 5 15 HNCO intensity (AU) - δ 125 0 115 20 40 60 80 100 PCNA e 30 130 PCNA–p15

T58 10 9 8 7 ) 20 –1

S72 δ 1 s ( N (p.p.m.) - H (p.p.m.)

15 N51 I65 - V53 R δ F68 W61 10 120 Y47 E67 c F69

E44 )

D75 Q62 –1 0 R70 0 K63 f 0.08 20 40 60 80 100 C54 –5

R14 L71 –10 0.06 K15 125 (p.p.m.) –15 A48 μ 0.04 Protein Kd ( M) N NH –20 p15 1.1 ± 0.1 0.93 ± 0.01 CSP 0.02 p1550–77 5.56 ± 0.03 1.04 ± 0.01 –25

Q per mol of injectant (kcal 0.00 8.6 8.4 8.2 8.0 0 1 2345 20 40 60 80 100 δ-1H (p.p.m.) Molar ratio p15 residue number

Figure 1 | The central region of p15 directly binds PCNA with low micromolar affinity. (a) Overlay of 1H-15N HSQC spectra of p15 (backbone amide region) in the absence (black) or presence (red) of PCNA at 1:4.2 molar ratio. Signals with strong intensity reduction on binding are labelled. (b) Overlay of 1H-15N TROSY spectra of PCNA in the absence (black) or presence (red) of p15 at 1:3.3 molar ratio. (c) ITC curves of p15 (black) and 50–77 p15 (green) binding to PCNA at 25 °C. The disociation constant (Kd) and stoichiometry (N) were derived from the fitting (lines) of the experimental data (squares) to a binding model assuming equivalent sites on PCNA. (d) NMR signal intensities from the 3D HNCO spectrum of p15 in the absence (black) or 15 presence (red) of PCNA. Absolute errors in intensities are less than 0.2% of the largest intensity in each spectrum. (e)Transverse NR2 relaxation rates of p15 in the absence (black) or presence (red) of PCNA measured at 25 °Cand101.4MHz.(f) CSP of p15 backbone amide signals (weighted average for 1HN and 15N) caused by the presence of PCNA. Dotted and dashed lines indicate the experimental error and average CSP plus one s.d., respectively. of free PCNA had disappeared (due to signal attenuation below (as named in the Protein Data Bank deposited coordinate file), the noise level or untraceable shifting; Supplementary Fig. 3). CSP which is therefore used for the structure discussion hereafter. projection onto the PCNA structure (Supplementary Fig. 4) The PIP-box residues of p1550–77 bind to the same groove on reveals that binding of all p15 constructs except the core fragment the front-face of the PCNA ring as the p21 PIP-box peptide7. The p1559–70 affects residues not only within the PIP-box-binding site side chain of Q62 is hydrogen bonded to PCNA-A252 and, via a on the front-face, but also on the inner side of the ring. water molecule, to PCNA-A208; residues I65-F67 form a short 310 helix, and the conserved hydrophobic trident formed by I65, F68 and F69 inserts into the hydrophobic cleft across the N- and Structure of the p15–PCNA complex. Our attempts to co- C-terminal domains of the PCNA protomer (Fig. 2c). Contrary to crystallize p15 and PCNA were unsuccessful and were most likely p21, the residues C-terminal to the p1550–77 PIP-box do not form hampered by the long flexible p15 termini. In contrast, truncated a stable intermolecular anti-parallel b-strand with the IDCL of p1550–77 co-crystallized with PCNA and diffracted up to 2.65 Å PCNA. Yet the major structural difference to p21 and other resolution (Table 1). Surprisingly, p1550–77 was found bound to PCNA-bound ligands involves residues N-terminal to the only two of the three PCNA protomers (Fig. 2a), while calori- p1550–77 PIP-box and concerns both the peptide conformation metry indicated 1:1 stoichiometry and the degenerate NMR and the interaction surface on PCNA. Residues P59-Q62 form a spectrum of PCNA showed equivalence of all three protomers type-I b-turn that reverses the peptide chain direction and sets both in the absence and presence of p15. This discrepancy arises the N-terminal residues N51-T58 to contact the inner surface of from the crystal packing, where a symmetry-related PCNA the PCNA ring where they are anchored to a groove between molecule occludes the peptide binding site on one protomer PCNA helices aA2 and aB2 through a network of intermolecular (Supplementary Fig. 5). Residues N51-L71 of p1550–77 are well contacts, including five direct and three water-mediated hydrogen defined in the crystal structure (Supplementary Fig. 6), although bonds (Fig. 2c and Supplementary Fig. 7). Contrarily, in the the side chains of N51, R70 and L71 could not be reliably crystal structures of other PIP-box peptides only a few of the modelled (Fig. 2b). The backbone conformation of the two bound residues N-terminal to the PIP-box are visible, and those seen peptides is very similar (Ca root mean squared deviation ¼ 0.32 Å), adopt a different conformation directing the chain away from the while the electron density map is better defined for chain D PCNA surface (Fig. 2d).

NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications 3 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439

Table 1 | X-ray diffraction data collection and refinement the binding site includes residues at the canonical PIP-box- statistics. binding groove on the front-face of the ring and extends to helices aA2 and aB2 at its inner surface. Moreover, several PCNA back- p1550–77–PCNA complex face residues also show significant signal attenuation and form a nearly contiguous region. These additional signal attenuations are Data collection not observed in the presence of substoichiometric amounts of the Space group P21 50–77 Unit cell parameters truncated p15 (Supplementary Fig. 9), and indicate a binding a, b, c (Å) 76.05, 41.23, 141.12 mode in which the full-length p15 N terminus passes though the a, b, g (°) 90, 105.39, 90 PCNA ring. Resolution (Å) 45.35–2.65 (2.79–2.65)* We then independently generated a large ensemble of p15– 50–77 Rmerge 0.093 (0.52) PCNA complex structures based only on the p15 –PCNA I/sI 9.3 (2.0) crystal structure with randomly appended p15 chain termini. In Completeness (%) 98.8 (93.9) 98% (92%) of these model complexes, at least one (two) p15 Redundancy 3.5 (3.0) chains pass through the ring. Only 13% of the complexes show all three p15 chains crossing through the ring, presumably due to Refinement Resolution (Å) 44.35–2.65 steric crowding inside the hole that results from building Number of reflections 24817 disordered chains randomly. Overall, these results indicate that the conformation of the tightly bound central region of p15 Rwork/Rfree 0.17/0.25 Number of atoms imposes a strong tendency on the p15 N-terminus to pass Protein 5941 through the ring. From this ensemble we selected structures Peptide 323 where N-terminal p15 residues lie close to the back-face PCNA Water 200 residues with significantly reduced NMR signal intensity. Ninety- r.m.s deviations two per cent of complex structures in this subset showed all three Bond lengths (Å) 0.016 p15 chains passing through the PCNA ring and emerging on its Bond angles (°) 1.786 back-face, while the remaining modelled complex structures had Average B factor (Å2) 49.91 at least one p15 chain fold back its N-terminus to the PCNA Ramachandran map front-face instead of funneling through the ring. A small set of Residues in favoured, allowed, 94.28, 3.86, 1.87 complex model structures illustrating these findings is shown in disallowed regions (%) Fig. 3b. Thus, while the NMR signal attenuations cannot be converted into exact populations of distinct p15 conformations, PCNA, proliferating cell nuclear antigen; r.m.s., root mean squared. *Values in parentheses are for highest-resolution shell. our results suggest that many (or most) of the p15 N-termini emerge at the back-face of PCNA. The models furthermore suggest that R41-E44 are those p15 residues that most likely The orientation of the p1550–77 peptide chain bound to PCNA contact the PCNA back-face. In line with this, the K42 NMR in our crystal structure suggests that, in solution, the N-terminal signal is attenuated beyond detection, A43 shows a significant 15 region of full-length p15 passes through the clamp channel to CSP and E44 a high NR2 rate in the presence of excess PCNA emerge on the back-face of the ring. This conclusion is consistent (Fig. 1d–f). with the NMR signal perturbations observed for residues at the inner side of the PCNA ring on p15 addition (Supplementary Fig. 4). We furthermore observed several residues on the back- Ubiquitin-independent proteasomal degradation of p15. The face with perturbed NMR frequencies and/or intensities. Some of ubiquitin ligase anaphase-promoting complex/cyclosome and its co-activator Cdh1 target p15 for proteasomal degradation them experience small frequency shifting along the titration with 10 p15 (like residue A194 in Supplementary Fig. 8a), while others do through its conserved KEN-box motif . This mechanism not shift but their intensities decay, vanishing at PCNA:p15 ratios involves recognition of polyubiquitin by the regulatory 19S cap of about 1:0.5 (like residues A82, E143 and K110 shown in subunits of the 26S proteasome and substrate unfolding by its Supplementary Fig. 8a) due to transient p15 contacts at the ATPase subunits before proteolysis within the chamber of the 20S PCNA back-side. Protein aggregation can be ruled out as an proteasome subcomplex. We found that p15 could also be alternative explanation for the signal decay, as confirmed by degraded by the 20S proteasome independent of ubiquitin and analytical size exclusion chromatography (Supplementary ATP, most likely due to its intrinsically disordered state. This Fig. 8b). Rapid NMR signal decay at substoichiometric ratios degradation is strongly and specifically inhibited in the presence without signal reappearance at ligand saturation has also been of PCNA (Fig. 4a), despite most of the p15 chain remaining reported for other protein complexes involving weak transient flexible and disordered in the bound form. The measured interactions between charged surfaces21 or a multitude of binding degradation rate is lower than calculated assuming that p15 can states22. For p15-bound PCNA at the end point of the titration, only be degraded when dissociated from PCNA (Fig. 4b). This we observe new signals in the vicinity of some of the vanished reduction indicates that p15 is less accessible to the proteasome ones, but signal broadening and attenuation in the p15–PCNA even when dissociating from PCNA, possibly because part of it complex made their unambiguous assignment impossible even remains trapped within the PCNA ring. with TROSY-based triple resonance experiments23. We therefore quantified PCNA signal intensities at substoichiometric DNA-binding activity of p15. PCNA is loaded on DNA by the PCNA:p15 ratios, where signal positions are still largely as in clamp loader, which binds, opens, and closes the ring under ATP the free form with selective intensity reduction for signals affected hydrolysis24. Although the inner surface of the ring is lined with by p15 binding. Since assignment of the NMR signals in this helices containing basic residues that complement the negative sample is unambiguous, we can confidently project signal charge of the DNA backbone, a tight PCNA–DNA interaction is attenuations onto the PCNA structure to derive the extended not expected as it would hinder sliding along the DNA. binding surface for full-length p15 (Fig. 3a). In agreement with Bacterial25 and archaeal26 clamps bind short DNA molecules, the crystal structure of PCNA bound to the p1550–77 fragment, but co-crystals of yeast PCNA and a short primed DNA duplex

4 NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439 ARTICLE

C N Ile 255 Pro 253 Lys 254 HN Ala 208 O

NH O Lys 60 HO Tyr 211 O Trp 61 IDCL Gln 62 Pro 59 OH Asp 156 C O H2N NH Thr 58 Ala 252 Lys 63 O N NH O A2 NH CO His 44 Pro 57 Arg 210 Gly 64 H2N Hydrophobic O CO NH pocket NH H2N N O Gly 155 Met 40 Arg 56 O B2 Val 45 Ile 65 Val 45 NH2 C Leu 47 NH Gly 66 NH Leu 126 NH2 Gly 127 Glu 67 O Val 55 Ile 128 Phe 68 Ser 152 Pro 129 Phe 69 CO Pro 234 90° Tyr 250 Arg 70 Cys 54 HO Leu 251 Ala 252 NH His 153 Pro 253 NH OC Val 53 HN Gly 127 C terminus CO Pro 52 Arg 149 ι Asn 213 p15 Pol Asn 51 p21 RFC p66 Fen1 N terminus

Figure 2 | Crystal structure of the p1550–77–PCNA complex. (a) Surface representation of one of the three p1550–77-binding sites on the PCNA ring, with 50–77 50–77 the bound p15 shown in sticks coloured by atom type. The omit 2F0–FC electron density map is displayed at 1.2s. N- and C-termini of p15 are indicated. (b) Ribbon diagram of one PCNA protomer (green) with bound p1550–77 (yellow). The IDCL and helices aA2 and aB2 (at the inner surface of the PCNA ring) are labelled. (c) Scheme of interactions between p1550–77 and PCNA. Indicated are the residue names (framed for PCNA), bound water molecules (blue circles), hydrogen bonds (dotted lines) and non-bonded interactions (arrows). The 310 helix and type-I b-turn structures in the bound p1550–77 are shaded blue. (d) Front and side views of the PCNA ring (grey surface) with superposed backbones of bound PIP-box fragments from six co-crystal structures: p1551–71 (yellow; PDB: 4D2G), p21143–160 (red; PDB: 1AXC), Pold p66453–465 (cyan; PDB: 1U76), Poli421–429 (blue; PDB: 2ZVM), yeast RFC389–406 (green; PDB:1SXJ) and Fen1336–348 (magenta; PDB:1U7B). Only the N-terminus of the p15 fragment is found to contact the inner surface of the ring. that binds to the bacterial clamp with nM affinity show only weak electrostatic (Supplementary Fig. 11b). The affinity of p15DN is features within the hole of the ring, interpreted as DNA being reduced by two orders of magnitude, indicating that the basic highly disordered or having low occupancy27. We have analysed N-terminal region of p15 contributes most to the DNA affinity the interaction of human PCNA with the same primed DNA in (Fig. 5a). The central fragment p1550–77 binds DNA with even solution by NMR. Even at a high DNA:PCNA molar ratio of 50:1, lower affinity, and no DNA binding was observed with the core only very small changes in the 1H-15N NMR signals of PCNA PIP-box fragment p1559–70. When DNA was titrated with PCNA- were observed, and affected residues do not localize to any bound p15, the increase in fluorescence polarization was larger specific area, but are sparsely distributed on the protein surface (Fig. 5a), indicating that a ternary complex forms (since the larger (Supplementary Fig. 10). This suggests that a direct DNA–PCNA the size of the complex the larger the polarization). The slightly interaction is either absent or extremely weak and unspecific. reduced DNA affinity shows that PCNA interferes only weakly Several clusters of basic residues in the N- and C-terminal with p15 binding to DNA. Similarly, when DNA-bound p15 was regions of p15 suggest that it may bind DNA through electrostatic titrated with PCNA (Fig. 5b) the PCNA affinity was comparable interactions. Our modelled structure of the p15–PCNA complex to that of isolated p15 (or p1550–77). Thus, a ternary p15–PCNA– likewise suggests that p15, when brought close to the DNA by DNA complex forms in solution in which the direct p15–PCNA binding to a loaded PCNA ring, may bind the DNA with its and p15–DNA interactions are largely independent of each other. flexible tails. To test the hypothesis of a p15–DNA interaction, The interaction of p15 with DNA was further characterized by we employed a fluorescence polarization assay and tested NMR using a 24 bp duplex (Supplementary Table 2). Perturba- DNA sequences reported to bind to S. solfataricus PCNA26 tions in several p15 resonances confirmed the binding (Fig. 5c). (Supplementary Table 2). The largest CSP occur near clusters of basic residues mainly in the In contrast to its archaeal homologue, human PCNA does not N-terminal half (Fig. 5d). Conversely, the smallest perturbations interact with these DNA sequences, while p15 does (Fig. 5). The occur near clusters of acidic residues in the C-terminal region. p15–DNA interaction is unspecific since any of the DNA Titration with DNA provokes linear shifts for several p15 NMR molecules tested (single-stranded DNA, primed DNA and signals in medium-salt buffer, but non-linear curved shifts in low- flapped DNA) are bound by p15 with similar affinity salt buffer (Supplementary Fig. 12). The latter indicates more (Supplementary Fig. 11a). The binding isotherms could be fitted than one binding site, while linear shifting is consistent with a to a 1:1 interaction model, with apparent dissociation constants single binding site28. These results confirm the electrostatic between 0.2 and 0.4 mMat25°C. Increasing the salt concentra- nature of the p15–DNA interaction with an affinity in the low tion from 40 to 150 mM lowers the affinity by an order of micromolar range at physiological ionic strength. The poor HN magnitude, indicating that the interaction is predominantly signal dispersion and only minor changes in Ca chemical shifts

NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications 5 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439

180°

90°

Figure 3 | Model of the full-length p15–PCNA complex. (a) Surface representation of PCNA (green) with bound p15 (yellow ribbon) showing the front (left) and back (right) face of the ring. The p1550–77–PCNA co-crystal structure with two bound peptide molecules was used as template and complemented by a third p15 peptide (in line with the ITC-derived stoichiometry). The locations of the disordered N- and C-termini are indicated by dashed ribbons. PCNA residues showing NMR signal attenuation in the presence of substoichiometric amounts of p15 (with relative TROSY signal intensity lower than the average minus one s.d.) are coloured red. PCNA residues whose NMR signal intensity could not be reliably measured and prolines are shown in grey. (b) Front and side views of 10 modelled p15–PCNA complex structures consistent with the experimental data. PCNA is shown as grey surface and p15 as ribbons, with the regions seen in the p1550–77–PCNA crystal structure coloured yellow and the added disordered N- and C-terminal extensions in red and blue, respectively. In one of the 10 selected complex models, one p15 N terminus folds back towards the front-face of PCNA instead of passing through the clamp, roughly representing the distribution found in the set of 5,000 modelled complexes.

(Supplementary Fig. 13) indicate that p15 does not acquire a These observations indicate that p15-bound DNA in the ternary defined structure on binding to the DNA; rather, its backbone complex is (at least transiently) located within the PCNA ring. remains flexible and disordered. In the presence of PCNA, DNA- The micrographs do not inform about the location of p15, and we bound p15 shows a similar pattern of NMR signal attenuation as cannot confirm whether p15 passes through the centre of PCNA in the absence of DNA (Fig. 5c), in line with a ternary complex also in the ternary complex. However, the high similarity in the where the central region of p15 is tightly bound to PCNA. pattern of p15 NMR signals that vanish when bound to PCNA To assess whether a DNA duplex could fit inside the p15- both in the presence and in the absence of DNA (Figs 1a and 5c) bound PCNA ring, we built models based on the crystal structure indicates that the structure of the tightly bound region is the of the p1550–77–PCNA complex with the p15 N- and C-termini same, with its N-terminus oriented towards the back-face, protruding from opposite faces of the PCNA ring. Figure 6a suggesting that it passes through the ring as in the binary shows that formation of such a complex is possible from steric p15–PCNA complex. and energetic considerations (Supplementary Fig. 14). For an experimental verification of this possibility, PCNA in the presence of p15, DNA or both was stained with uranyl acetate Discussion and observed by transmission electron microscopy (Fig. 6b). We have used an integrative structural approach to characterize Two-dimensional analysis of single particle projections yielded the interaction of the intrinsically disordered protein p15 with the class averages for top-, side-, and tilted-views of the PCNA ring, DNA-sliding clamp PCNA. By NMR, we identified the p15 core with a larger proportion of tilted-view classes for the ternary p15– interaction region and designed a p1550–77 fragment with DNA–PCNA complex (Supplementary Fig. 15). Most remark- truncation of the flexible tails that impeded crystallization ably, many of the top- and tilted-views of the ternary complex of the full-length p15–PCNA complex. The well-diffracting show increased density within the PCNA ring compared to both p1550–77–PCNA crystals reveal interactions also outside the binary mixtures (where only the p15–PCNA mixture, however, is canonical PIP-box that reach into the PCNA ring and expected to form a complex). The same result is obtained when substantially contribute to the binding free energy. ITC the rotational power spectra of individual particles in the three measurements also confirm that one p15 molecule binds to each data sets are calculated and classified (Supplementary Fig. 16). of the three protomers in the PCNA ring with low mM affinity,

6 NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439 ARTICLE

Proteins in reaction mixture hierarchy based on relative affinities modulated by post- p15 + + + + + + + + + + translational modifications that, together with steric hindrance and relative protein concentrations, determine the occupancy of 20S – + + + – + + + – + 30 PCNA – – – – + + + + – – the PCNA protomers . The simultaneous occupation of all three BSA – – – – – – – – + + sites by identical ligands, as seen here and in other quantitative kDa M binding studies, may not be the prevalent situation inside the cell. 66 Only the small and disordered p21 protein might have an affinity BSA high enough to displace any other ligand from PCNA, explaining 45 its dominant inhibition of DNA replication and cell cycle 31 32 35 progression . For instance, a C-terminal p21 fragment binds PCNA PCNA with two orders of magnitude higher affinity than p15. 25 20S Conversely, the affinity of PCNA for p15 is higher than for a PIP- 18 box fragment of the p66 subunit of the replicative polymerase d p15 29 14 (Pold; Kd ¼ 15.4 mM, measured by ITC at 30 °C) , and similar to the translesion synthesis polymerase Z (TLS PolZ; Kd ¼ 0.4 mM, 33 0 124012404 measured by surface plasmon resonance at 25 °C) . Pold is a four-subunit complex of 239 kDa, much larger than PCNA, Reaction time (h) which possibly occludes (by binding and steric hindrance) more 1.0 than one PIP-box-binding groove, as observed in the complexes Sample k (s–1.10–4) deg of large proteins with PCNA rings34,35. PolZ is a 78-kDa p15 5.48 ± 0.06 multidomain protein with an ubiquitin-binding domain that 0.8 p15–PCNA 0.32 ± 0.01 increases its affinity for PCNA when ubiquitylated at K164 after 36

0 ultraviolet irradiation . 0.6 The rapid degradation of p15 by the 20S proteasome core particle suggests that this may be a mechanism for regulating free 0.4 p15 levels in the cell, in addition to the ubiquitin- and ATP- [p15] / dependent one. PCNA binding, however, strongly inhibits p15 degradation by the 20S proteasome, and the degradation rate is 0.2 lower than expected if dissociation from PCNA was not rate limiting. Apparently, p15 is less accessible to the proteasome even 0.0 after dissociation from PCNA, consistent with the modelled 0231 45 complex structure where p15 may stay partly trapped within the Reaction time (h) PCNA ring. Degradation by the 20S proteasome and its inhibition on binding to PCNA have also been described for p21 (ref. 37). Figure 4 | Ubiquitin- and ATP-independent degradation of p15 by the A short sequence C-terminal to the p21 PIP-box is essential for 20S proteasome. (a) SDS–PAGE of different reaction mixtures and times, binding to one of the a-subunits of the 20S proteasome. Part of this as indicated. The left lane contains the molecular weight markers. sequence (156RLIFSK161) is tightly bound to PCNA in the crystal (b) Densitometric analysis of SDS–PAGE stains showing the p15 structure of the complex and PCNA binding might mask the degradation by the 20S proteasome. Filled and open circles show the recognition site. The corresponding sequence, however, is very fraction of p15 remaining after the indicated incubation times with 20S different in p15 (74KDSEKE79) and is not tightly bound to proteasome in the absence and presence of fivefold molar excess of PCNA, PCNA. respectively. The data were averaged over two independent experiments Both p15 and PCNA are primarily expressed during the and fitted to a monoexponential decay (solid lines). The dashed line S-phase of the cell cycle, when chromosomal replication occurs, represents the fraction of remaining p15 calculated assuming that its and both are mainly associated with the chromatin10,11. Thus, dissociation from PCNA is not rate limiting for degradation, with the while p15 may be bound to PCNA alone, its functionally relevant maximal variation (from estimated experimental uncertainty, see Methods) form in DNA replication and repair is that bound to the indicated in grey. The values of the degradation rate kdeg considering this À 4 À 1 DNA-loaded PCNA. We have found that p15 by itself binds to maximal variation are between 0.55 and 0.82 Á 10 s . DNA, predominantly via electrostatic interactions with its basic N-terminal tail. This region contains an 21APRK24 sequence although our crystal structure shows only two bound p1550–77 similar to the known DNA-binding motif SP(R/K)(R/K)38 that is molecules due to a crystal packing artifact. The non-cooperative present in p15 homologues as well as in the N-terminal histone 1:1 stoichiometry is common to all PIP-box containing proteins H3 tail (Supplementary Fig. 17). The flexible N-terminal histone whose binding to PCNA has been studied by calorimetry19,29. tails are likewise rich in basic residues and have been proposed to The mode of p15 binding to PCNA, however, differs importantly unspecifically bind DNA, thus attenuating electrostatic repulsion in that p15 extends its interaction site N-terminally from the in the chromatin39. PIP-box-binding groove on the front-face to the inner surface of While human PCNA alone does not interact noticeably with the PCNA ring. Both the backbone orientation of the p1550–77 DNA, p15 can bind to PCNA and DNA simultaneously via N terminus visible in the crystal structure and the NMR data with independent sites and with similar affinities. Our structure-based full-length p15 strongly suggest that the flexible N-terminus model of the ternary complex shows that the PCNA trimer may passes through the PCNA ring to emerge, and transiently contact, accommodate up to three p15 molecules plus the duplex DNA on its back-face. In the ensemble of p15–PCNA complexes within its central cavity (Fig. 6a). This model is consistent with modelled from these findings, most of the bound p15 chains have our electron micrographs that show the PCNA ring at least their flexible N- and C-termini protrude at the opposite back and partially filled by p15-bound DNA (Fig. 6b). On the basis of these front faces of the PCNA ring, respectively (Fig. 3b). observations, we propose that the flexible N-terminal tail of p15 In the nucleus of the cell, the various PCNA ligands compete may act as an extended arm that captures and binds the DNA for binding to the three protomers of the ring and there likely is a fibre at the back-face of PCNA. It could, thus, modulate and slow

NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications 7 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439

350 μ Protein Kd ( M) 300 p15–PCNA 0.38 ± 0.04 G49 p15 0.19 ± 0.01 G50 Δ 250 p15 N 22 ± 7 110 G64 p1550–77 52 ± 7 200 p1559–70 ND G66 PCNA ND 150 p15 p15–DNA p15–PCNA–DNA 100 115 50 Fluorescence polarization (mP)

0 T58 0 0.001 0.01 0.1 1 10 100 1,000

[Protein] (μM) N (p.p.m.) S72

15 I65 - W61 350 δ Y47 V53 Sample K (μM) N51 F68 d 120 E44 E67 F69 DNA–p15 2 ± 1 Q62 300 DNA–p1550–77 10 ± 2 D75 K63

C54 250 R70 L71 R14 200 125 A48

150 Fluorescence polarization (mP)

100 0 0.001 0.01 0.1 1 10 100 1,000 8.6 8.4 8.2 8.0 [PCNA] (μM) δ-1H (p.p.m.)

0.15

0.10 (p.p.m.)

NH 0.05 CSP 0.00 10 20 30 40 50 60 70 80 90 100 110 p15 residue number Figure 5 | DNA-binding activity of p15. (a) Fluorescence polarization of 50 nM fluorescein-labelled flapped DNA in the presence of increasing concentrations of different p15 constructs, PCNA, or a 1:3 (molar ratio) mixture of p15 and PCNA. Measurements (filled circles) were averaged over three independent experiments, with the s.d. indicated by the error bars, and fitted to a binding model assuming 1:1 stoichiometry (lines). (b) Fluorescence polarization of flapped DNA in the presence of 6 mM p15 (black) or 124 mM p1550–77 (red), and increasing PCNA concentration. Data analysis was done as in a.(c) Overlay of 1H-15N HSQC spectra (backbone amide region) of p15 alone (black), in the presence of dsDNA (red) and in the presence of both dsDNA and PCNA (blue). The three samples contained low salt buffer. (d) CSP of p15 backbone amide signals caused by dsDNA binding in low salt buffer. Dotted and dashed lines indicate the experimental error and average CSP plus one s.d., respectively. The amino acid sequence of p15 is indicated above the plot. Residues with positive charge are coloured blue, negative red and neutral grey. All data were measured at 25 °C. down the linear diffusion of PCNA along the DNA (Fig. 6c). damage response to prevent rapid drift of PCNA at stalled Since the mean diffusion coefficient of human PCNA sliding replication forks between DNA polymerase swapping events. along dsDNA is D ¼ 1.16 mm2 s À 1, unrestrained PCNA would Interestingly, a fraction of PCNA-bound p15 is ubiquitylated at diffuse by several hundred bp in a few ms40. K15 and K24 during normal S-phase progression and is The E. coli b-clamp binds DNA with nanomolar affinity and selectively degraded by the proteasome after ultraviolet has a coefficient of diffusion on DNA that is two orders of irradiation11. The affinity of p15 for DNA would likely be magnitude lower than for human PCNA41. The processivity reduced by ubiquitylation of these lysines in the p15 N-terminal subunit of the herpes simplex virus DNA polymerase UL42 region that primarily interacts with the DNA, as it removes likewise has nanomolar affinity for DNA, which is remarkably positive charges and introduces steric hindrance. Thus, ultraviolet high for a sequence-independent DNA-binding protein42, while irradiation would shift the equilibrium from ubiquitylated to the interaction of S. cerevisiae PCNA with DNA is very weak27. non-ubiquitylated PCNA-bound p15 that is more competent for This comparison suggests a loss in DNA affinity during the DNA binding, and the switch from the replicative polymerases evolution of DNA-sliding clamps, with a concomitant gain in (like Pold) to TLS polymerases (like PolZ) could be facilitated. processivity that implies faster diffusion along the DNA. In Our proposed model for p15-modulated PCNA sliding along higher organisms, p15 might help to regulate the sliding velocity the DNA could in future be validated by single-molecule of PCNA, and this p15 function may be required for the DNA experiments40.

8 NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439 ARTICLE

p15–PCNA

90°

p15–PCNA–DNA

Replication Replication Polδ p12 p12 PCNA PCNA–DNA p66 p66 p125 p125 DNA N p50 p50

Ub p15 N Ub Ubiquitin C C

Figure 6 | Model of the ternary p15–PCNA–DNA complex. (a) Ribbon plot of PCNA (green), p15 (yellow) and a regular B-form 24 bp DNA duplex (orange and blue). (b) Electron microscopy pictures of the mixtures with negative staining. Representative areas of digital micrographs are shown at the left, selected class averages resulting from 2D reference-free alignment at the right (top-, tilted- and side-views from top to bottom). The bars drawn on each of the three micrographs are 100 nm long. (c) Scheme for a possible arrangement of the PCNA ring sliding along the DNA, with the replication polymerase d complex bound at the front-face. In this model, one p15 chain traverses the clamp channel, with its N-terminal tail binding the DNA at the back-face of PCNA. This p15–DNA interaction could be disrupted by p15 ubiquitylation at K15 and K24 occurring during S-phase progression and, thus, modulate PCNA diffusion along the DNA.

The PIP-box/PCNA interaction is a promising target for cancer 200 26/60 column equilibrated with PBS, pH 7.0, and then exchanged into the therapeutics43,44. Yet, PCNA directs fundamental DNA- crystallization buffer (20 mM Tris-HCl, pH 7.5, 10% glycerol, 2 mM dithiothreitol templated processes, and targeting its common interaction site (DTT)) using a PD10 column. Stock solutions in PBS or crystallization buffer were flash-frozen in liquid nitrogen and stored at À 80 °C. The protein concentrations for so many different PIP-box proteins might cause unspecific were measured by absorbance at 280 nm using the extinction coefficient calculated cytotoxicity. Our studies reveal that p15 exhibits a so-far unique from the amino acid composition (15,930 M À 1 cm À 1). All indicated binding mode among PIP-box proteins in which residues concentrations of PCNA samples refer to protomer concentrations. N-terminal to the PIP-box contribute importantly by binding Human p15 (Uniprot: Q15004) was produced using E. coli BL21(DE3) cells grown in the appropriate culture media and purified by three sequential on the inside of the PCNA ring. Targeting this novel interaction chromatographic separations: anion exchange, reverse phase, and cation might provide an avenue to inhibitors with high selectivity for the exchange17. This purified protein does not contain any extraneous residue, but oncogenic p15. mass spectrometry showed that it lacks the initial methionine. The gene encoding the p15 mutant with residues 2–31 deleted (p15DN) was generated using the QuikChange Site-Directed Mutagenesis Kit (Agilent). The p15DN–pET11d Methods construct was introduced in E. coli BL21(DE3) cells and grown at 37 °C until Clones and proteins. Human PCNA (UniProt: P12004) was produced in E. coli OD600 ¼ 0.6, when protein expression was induced with 1 mM isopropyl-b-D- BL21(DE3) cells grown in appropriate culture media to obtain protein with natural thiogalactoside for 3 h at 37 °C. Cells were harvested and resuspended in lysis isotopic abundance or uniform enrichment using a clone with N-terminal His6-tag buffer (20 mM Tris, pH 8.0, 1 mM DTT, 1 mM EDTA plus one tablet of the protein and PreScission protease cleavage site in a pET-derived plasmid. For NMR inhibitor cocktail Complete from Roche) and sonicated on ice. Cells were pelleted samples, the protein was purified from the soluble fraction by Co2 þ -affinity by centrifugation at 4 °C at 25,000 r.p.m. for 1 h. After centrifugation, the chromatography, cleaved by PreScission protease, and polished by gel filtration supernatant was loaded on a Hiload 26/10 Q Sepharose HP column. The column chromatography45. The NMR sample buffer was PBS (137 mM NaCl, 2.7 mM KCl, was washed with 20 mM Tris, pH 8.0, 1 mM DTT, 1 mM EDTA (Buffer A) and the 10 mM sodium phosphate, 2 mM potassium phosphate) pH 7.0. All columns and protein was eluted with a two-step gradient using buffer A with 2 M NaCl (0–20% chromatography systems were from GE Healthcare. Protein elution was monitored in 0.3 CV and 20–100% in 1 CV). Fractions containing the protein, as seen in by absorbance at 280 nm and confirmed by SDS–polyacrylamide gel SDS–PAGE, were pooled, diluted three times in buffer A, and loaded on a HiLoad electrophoresis (SDS–PAGE). The purified protein contained the extra sequence 26/10 SP Sepharose HP column. The protein was eluted with a 0–51% gradient of GSH at the N-terminus. The PCNA sample for crystallization was obtained by buffer B and the fractions containing the protein were concentrated and polished introducing two additional purification steps. The sample cleaved with PreScission by gel filtration on a Superdex 75 26/60 column previously equilibrated with PBS, protease was dialysed against 50 mM sodium acetate pH 5.5, 100 mM NaCl. After pH 7.0, 1 mM DTT. Mass spectrometry analysis of the purified material confirmed separation of some precipitated material, the solution was loaded on a HiTrap the identity of the polypeptide, which lacked the initial methionine of the construct Heparin HP column equilibrated with the same buffer. After column washing, the and contained residues 32–111 of human p15. protein was eluted with a 0–100% gradient of 50 mM sodium acetate pH 5.5, 2 M Peptides p1550–77 (50GNPVCVRPTPKWQKGIGEFFRLSPKDSE77)and NaCl in 20 column volumes (CV). The protein-containing fractions of the major p1559–70 were purchased from Apeptide Co. For NMR, ITC and fluorescence peak were dialysed against 20 mM Tris-HCl buffer pH 7.6, 150 mM NaCl and polarization measurements, concentrated peptide stock solutions were prepared injected into a HiTrap Chelating HP column loaded with Co2 þ cations to remove dissolving the lyophilized powder in water or in PBS, pH 7.0, and the pH was adjusted uncleaved PCNA. The flow-through was loaded on a HiTrap Q Sepharose column with NaOH, if necessary. For crystallization, the buffer was 20 mM Tris, pH 7.5, and eluted with a 0–60% gradient of 20 mM Tris-HCl pH 7.6, 1 M NaCl in 5 CV. 0.5 mM TCEP. The peptide concentration was measured by absorbance at 280 nm The protein-containing fractions were concentrated and polished using a Superdex using the extinction coefficient calculated from the amino acid composition.

NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications 9 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439

NMR spectroscopy. Unless indicated otherwise, all NMR data were measured on Fluorescence spectroscopy. Deoxyoligonucleotides for fluorescence polarization a Bruker Avance III 800 MHz (18.8 T) spectrometer equipped with a cryogenically experiments were purchased from Thermo Fischer Scientific, and their con- cooled triple resonance z-gradient probe. 1H-15N TROSY spectra of PCNA were centration was determined from the extinction coefficient at 260 nm. The DNA recorded at 35 °C on samples of 50 mMU-[2H,15N] PCNA with or without p15 molecules were obtained by mixing equimolar amounts of the appropriate oligo- molecules at different concentrations in PBS pH 7.0, 1 mM DTT. The spectra used nucleotides (indicated in Supplementary Table 2) in 20 mM Tris-HCl, pH 7.8, for assignment of p15 bound to PCNA were recorded at 25 °C on a sample con- 25 mM NaCl, at 93 °C for 2 min with subsequent slow cooling at room tempera- taining 0.15 mM U-[13C,15N] p15, 0.75 mM U-[2H] PCNA in PBS, pH 6.9, 1 mM ture. Binding experiments used samples with 50 nM fluorescein-labelled DNA in N 0 a b À 1 DTT, 5% D2O. Assignment of H ,N,C,C and C resonances was obtained from 30 mM HEPES, pH 7.6, 40 mM KCl, 5% glycerol, 0.1 g l bovine serum albumin, three-dimensional (3D) HNCO, HN(CA)CO, HNCA, HN(CO)CA, HNCACB and 2 mM TCEP with increasing concentrations of the different p15 molecules or HN(CO)CACB spectra, assisted by the previously published free p15 assignment17. PCNA. Fluorescence polarization measurements were performed at 25 °Cona 1H chemical shifts were referenced to internal 2,2-dimethyl-2-silapentane-5- FluoroMax-3 spectrometer (Horiba Scientific) with a 10 mm Hellma quartz cuvette sulfonate (DSS, 0.00 p.p.m.), and 13C and 15N chemical shifts were indirectly using the constant wavelength module for quantitative analysis. The excitation 46 15 referenced to DSS . The p15 backbone NR2 relaxation rates at 25 °C and wavelength was 490 nm and the emission wavelength 515 nm. s.d. were calculated 101.4 MHz were measured on a Bruker Avance 1 GHz (23.4 T) spectrometer from the polarization values of three repeated measurements. Kaleidagraph equipped with a cryogenically cooled triple resonance z-gradient probe. A series of (Synergy Software) was used to fit the experimental data to a one-site-binding two-dimensional 1H-15N correlation maps modulated by different values of the model. 15N transverse relaxation delay were acquired in a fully interleaved mode17.The samples contained 180 mMU-[15N] p15 in 200 mM MES buffer, pH 6.8, 0.5 mM Molecular modelling. The structure of full-length p15 bound to PCNA was DTT, 5% D2O with or without 750 mM unlabelled PCNA. These two samples were 50–77 prepared from stocks of p15 and PCNA simultaneously dialysed in the same beaker modelled starting from the crystal structure of the p15 –PCNA complex. The conformation of residues 51 to 71 was maintained as in the crystal structure, and and buffer, eliminating the possibility for bias in the measured CSP (shown in disordered N- and C-terminal extensions were built into each p1550–77 peptide Fig. 1f) by differences in pH or ionic strength. CSP values and their errors were using Flexible-Meccano54 to sequentially generate peptide chains based on amino computed as the weighted average distance between backbone amide 1H and 15N acid specific conformational propensity and volume exclusion with statistical chemical shifts in the free and bound states47. NMR titration of PCNA with p15 conformation sampling. An ensemble of 100,000 full-length p15 conformers was was performed using stocks of 50 mM 2H,15N-PCNA and 1 mM p15 in PBS, NaN 3 generated for each of the three bound peptides, and side chains were added to each 0.02%, pH 7.3 recording 1H-15N TROSY spectra at 35 °C and different PCNA:p15 conformer using SCCOMP55. These p15 conformers were then docked individually ratios. Quantitative signal intensity analysis at 1:0.2 ratio was done on 200 mM PCNA samples with 40 mM p15 or p1550–77. NMR signals of p15 samples at onto the PIP-box binding site of PCNA and discarded if steric clashes with PCNA were detected, resulting in 11,000 p15 conformers for each of the three equivalent concentrations between 20 mM and 1 mM did not show any sign of aggregation. binding sites on the PCNA trimer. These conformers were then randomly selected Analytical size exclusion chromatography of 50 mM PCNA, 250 mM p15, or the and docked on the PCNA ring to generate 5,000 models of the complex with three mixture of both, performed at room temperature on a Superdex 200 bound p15 chains and no steric clashes between them. Another set of 5,000 models 10/300 GL column, showed, homogenous samples with no indication of of the complex with three bound p15 chains was generated in the same way but aggregation. using subsets of the p15 conformers with at least one residue in their N-terminal PCNA binding to the short primed DNA was investigated at 35 °Cby tails (V2-G50) closer than 6.5 Å (Cb À Cb distance) to any of those PCNA comparison of 1H-15N TROSY spectra recorded on a 50-mMU-[15N,2H] PCNA in PBS, pH 7.0, with or without 2.5 mM DNA. NMR titrations of p15 with dsDNA back-face residues showing reduced TROSY signal intensity in the presence of substoichiometric amounts of p15 (that is, PCNA residues V7, A82, E85, F103, in PBS or MES buffer were performed by adding small volumes of highly K110 and E143). To classify the p15 chains as crossing the ring or not, the centre concentrated DNA stock solutions to p15 samples, such that changes in p15 of mass of PCNA and a plane passing through this centre and perpendicular to the concentration remained small. Assignment of backbone HN,Ha and N three-fold symmetry axis of PCNA were defined. A p15 chain was considered chemical shifts of p15 bound to dsDNA in low-salt buffer (20 mM MES, 1 mM to go through PCNA if at least one of the heavy atoms in residues 2–50 fulfils both DTT, pH 6.1) at 25 °C was obtained from 3D 1H-15N HSQC-TOCSY, 1H-15N of the following two criteria: (i) the vector connecting the centre of mass of PCNA HSQC-NOESY and HNHA experiments recorded on a sample containing 30 mM and the heavy atom has a positive dot product with the normal to the plane U-[15N] p15 and 42 mM dsDNA. Assignment of the HN,N,C0,Ca and Cb directed to the back-face: (ii) the distance between the heavy atom and the PCNA resonances of p15 bound to dsDNA in medium-salt buffer (PBS, pH 6.3, 1 mM centre of mass is less than 17 Å (half the diameter of the PCNA ring; to exclude DTT) at 25 °C was obtained from 3D HNCO, HN(CO)CA, HNCA and chains that comply with the first criterion but going over the ring instead of HN(CO)CACB experiments recorded on a sample containing 50 mMU-[13C,15N] p15 and 125 mM dsDNA. through it). To model the ternary p15–PCNA–DNA complex, disordered tails of p15 were added to the p1550–77–PCNA crystal structure using Coot51, where backbone dihedral angles were adjusted manually to maximize the empty space within the Isothermal titration calorimetry. These measurements employed an ITC200 PCNA central cavity. The 24 bp DNA duplex in regular B conformation was then calorimeter. Protein solutions (dialysed against PBS, pH 7.0, 2 mM TCEP) were placed within the channel, and the whole ternary complex structure was refined by simulated annealing and energy minimization using loose and ambiguous distance titrated with concentrated peptide stocks prepared by dissolving the lyophilized 56 peptides in the dialysis buffer and adjusting the pH to 7.0 with NaOH, if necessary. restraints between DNA and PCNA in the HADDOCK refinement protocol . The binding isotherms were analysed by non-linear least-squares fitting of the Atoms at the interface (within 6.5 Å) were allowed to sample additional experimental data to a model assuming a single set of equivalent sites48. conformations during a semi-flexible simulated annealing, and further refinement Data analysis employed Microcal Origin (OriginLab) and in-house developed in explicit water solvent was performed. software. Proteasomal degradation assays. Samples containing 1 mg p15, in the absence or presence of PCNA at 1:5 molar ratio (12.5 mg) or BSA at 2.2 molar ratio (12.4 mg, Sigma-Aldrich) were incubated with 2 mg human 20S proteasome subcomplex Protein crystallization and structure determination . Stocks of PCNA and (Enzo Life Sciences) at 37 °Cin16ml of 30 mM Tris-HCl, pH 7.6, 100 mM NaCl, p1550–77 solutions were mixed to final concentrations of 0.5 mM and 0.6 mM, 10 mM CaCl2, 2 mM MgCl2, 1 mM DTT, 5% (v/v) glycerol. The reaction was respectively (1:1.2 monomer molar ratio) and incubated at room temperature for stopped after 1, 2 or 4 h by adding gel loading buffer and boiling. Samples were run 30 min before screening crystallization conditions by the hanging drop vapour on a precast SDS–PAGE gel (4–12% gradient, Invitrogen) and stained with Coo- m diffusion method. Best diffracting co-crystals grew within 3 days at 18 °Cin2 l massie Brilliant Blue. The remaining p15 was quantified by densitometry. m m droplets obtained by mixing 1 l of the complex solution and 1 l of a solution Assuming that p15 can only be degraded after dissociating from PCNA, the containing 40% polyethylene glycol 300 in 0.1 M sodium phosphate-citrate buffer degradation mechanism can be described by a simple two-step reaction pH 4.2. Microseeding with seed beads (Hampton Research) was used to improve (dissociation from PCNA, then degradation): the crystals, which were flash-frozen directly from the drops. All data were collected at 100 K using synchrotron radiation (l ¼ 1.00 Å) at the koff kdeg PCNA À p15 , PCNA þ p15 À! p15deg PXI-XS06A beamline (SLS, Villigen, Switzerland) and a Pilatus 6M detector. kon Diffraction data processing and scaling was accomplished with X-ray Detector 49 50–77 The maximal degradation rate is given by the measured first-order rate Software (XDS) . The structure of the p15 –PCNA complex was determined À 4 À 1 by molecular replacement using Phaser50 and the previously reported human constant for free p15 (kdeg ¼ 5.5 Â 10 s , Fig. 4b). Degradation of p15 PCNA structure as a search model (PDB:1VYM). The initial model was positioned may be assumed to be the rate-limiting step, considering that the dissociation manually with Coot51 and refined using PHENIX52. Refinement and data constant is in the low micromolar range and that the binding to PCNA does collection statistics are summarized in Table 1. Protein–DNA hydrogen bonds and not induce major structural rearrangements in either partner. For comparison, Van der Waals contacts were identified and analyzed with the Protein, Interfaces, the dissociation rate constant of the p21–PCNA complex measured by surface À 4 À 1 ° Surfaces and Assemblies service (PDBePISA) at the European Bioinformatics plasmon resonance is koff ¼ 6.2 Â 10 s at 37 C (ref. 57). Taking into Institute53. account that the affinity of p21 for PCNA exceeds that of p15 by three orders of magnitude and that the stoichiometry of binding to PCNA is 1:1

10 NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications & 2015 Macmillan Publishers Limited. All rights reserved. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439 ARTICLE

for both, it can be expected that koff of p15 largely exceeds its intrinsic kdeg. 17. De Biasio, A. et al. p15(PAF) is an intrinsically disordered protein with With this assumption, the concentration of p15 that remains non-degraded nonrandom structural preferences at sites of interaction with other proteins. by the proteasome at time t can be calculated from the following Biophys. J. 106, 865–874 (2014). equation: 18. De Biasio, A. et al. Proliferating cell nuclear antigen (PCNA) interactions in solution studied by NMR. PLoS ONE 7, e48390 (2012). Zt 19. Zheleva, D. I. et al. A quantitative study of the in vitro binding of the ½p15 ðÞ¼t ½p15 0 À kdeg Á ½p15 freeðtÞÁdt C-terminal domain of p21 to PCNA: affinity, stoichiometry, and 0 thermodynamics. Biochemistry 39, 7388–7397 (2000). 20. Sanchez, R., Torres, D., Prieto, J., Blanco, F. J. & Campos-Olivas, R. Backbone assignment of human proliferating cell nuclear antigen. Biomol. NMR Assign. 1, where [p15]0 is the initial concentration of p15 and [p15]free is the concentration of free p15, which can be calculated using the dissociation constant at 37 °C 245–247 (2007). (Kd ¼ 3.3 mM). The equation was numerically solved assuming that [p15]free is 21. McAlister, M. S. et al. NMR analysis of interacting soluble forms of the cell-cell invariant within infinitesimally small intervals and the reaction follows a zero order recognition molecules CD2 and CD48. Biochemistry 35, 5982–5991 (1996). kinetics, where [p15] decreases linearly with time. The remaining p15 at each 22. Park, S. J., Kostic, M. & Dyson, H. J. Dynamic interaction of Hsp90 with its reaction time was derived by iteratively recalculating [p15]0 and [p15]free in steps of client protein p53. J. Mol. Biol. 411, 158–173 (2011). 1 s (smaller time intervals yielded the same protein concentrations within o0.1% 23. Klein, C. et al. NMR spectroscopy reveals the solution dimerization interface deviation). Figure 4b shows the calculated degradation curve together with the of p53 core domains bound to their consensus DNA. J. Biol. Chem. 276, maximal uncertainty derived from the estimated errors in protein concentration 49020–49027 (2001). (10%) and dissociation constant (10%), and the fitting error in the degradation rate 24. Jeruzalmi, D., O’Donnell, M. & Kuriyan, J. Clamp loaders and sliding clamps. À 4 À 1 of free p15 (0.06 Â 10 s ). Curr. Opin. Struct. Biol. 12, 217–224 (2002). 25. Georgescu, R. E. et al. Structure of a sliding clamp on DNA. Cell 132, 43–54 Electron microscopy. PCNA at 0.1–0.025 g l À 1 was mixed with a 1:6 molar excess (2008). of p15, 1:12 molar excess of dsDNA, or both (in PBS, pH 7.0, 1 mM DTT). The 26. Hutton, R. D., Craggs, T. D., White, M. F. & Penedo, J. C. PCNA and XPF samples were applied to glow-discharged carbon-coated copper grids and stained cooperate to distort DNA substrates. Nucleic Acids Res. 38, 1664–1675 with 1% (w/v) uranyl formate. Micrographs were taken at low radiation dose on a (2010). Jeol JEM-1230 LaB6 transmission electron microscope operated at 100 kV and 27. McNally, R., Bowman, G. D., Goedken, E. R., O’Donnell, M. & Kuriyan, J. equipped with an Orius SC1000 CCD camera. More than 100 micrographs with a Analysis of the role of PCNA-DNA contacts during clamp loading. BMC Struct. nominal magnification of 30,000 (2.3 Å per pixel) were recorded per sample. Biol. 10, 3 (2010). Sets of 8,759, 4,277 and 5,343 particles of the p15–PCNA–DNA, p15–PCNA and 28. Zuiderweg, E. R. Mapping protein-protein interactions in solution by NMR PCNA–DNA samples, respectively, were manually selected and windowed using spectroscopy. Biochemistry 41, 1–7 (2002). XMIPP58. These sets were subjected to an iterative process of reference-free 29. Bruning, J. B. & Shamoo, Y. Structural and thermodynamic analysis of human two-dimensional alignment and classification using the CL2D algorithm of XMIPP. PCNA with peptides derived from DNA polymerase-delta p66 subunit and flap An independent analysis was done for each of the three data sets based on the endonuclease-1. Structure 12, 2209–2219 (2004). rotational power spectra of each individual particle, which were classified using a 30. Mailand, N., Gibbs-Seymour, I. & Bekker-Jensen, S. Regulation of 7 Â 7or10Â 10 Kernel probability Density Estimator Self-Organizing Map 59 PCNA-protein interactions for genome stability. Nat. Rev. Mol. Cell Biol. 14, (KerDenSOM) classificatory matrix . 269–282 (2013). 31. Abbas, T. & Dutta, A. p21 in cancer: intricate networks and multiple activities. References Nat. Rev. Cancer 9, 400–414 (2009). 1. Moldovan, G. L., Pfander, B. & Jentsch, S. PCNA, the maestro of the replication 32. Chen, J. et al. A 39 amino acid fragment of the cell cycle regulator p21 is fork. Cell 129, 665–679 (2007). sufficient to bind PCNA and partially inhibit DNA replication in vivo. Nucleic 2. Krishna, T. S., Kong, X. P., Gary, S., Burgers, P. M. & Kuriyan, J. Crystal Acids Res. 24, 1727–1733 (1996). structure of the eukaryotic DNA polymerase processivity factor PCNA. Cell 79, 33. Hishiki, A. et al. Structural basis for novel interactions between human 1233–1243 (1994). translesion synthesis polymerases and proliferating cell nuclear antigen. J. Biol. 3. Wing, R. et al. Crystal structure analysis of a complete turn of B-DNA. Nature Chem. 284, 10552–10560 (2009). 287, 755–758 (1980). 34. Bowman, G. D., O’Donnell, M. & Kuriyan, J. Structural analysis of a eukaryotic 4. Maga, G. & Hubscher, U. Proliferating cell nuclear antigen (PCNA): a dancer sliding DNA clamp-clamp loader complex. Nature 429, 724–730 (2004). with many partners. J. Cell Sci. 116, 3051–3060 (2003). 35. Bubeck, D. et al. PCNA directs type 2 RNase H activity on DNA replication and 5. De Biasio, A. & Blanco, F. J. Proliferating cell nuclear antigen structure and repair substrates. Nucleic Acids Res. 39, 3652–3666 (2011). interactions: too many partners for one dancer? Adv. Protein Chem. Struct. Biol. 36. Freudenthal, B. D., Gakhar, L., Ramaswamy, S. & Washington, M. T. Structure 91, 1–36 (2013). of monoubiquitinated PCNA and implications for translesion synthesis and 6. Jonsson, Z. O., Hindges, R. & Hubscher, U. Regulation of DNA replication and DNA polymerase exchange. Nat. Struct. Mol. Biol. 17, 479–484 (2010). repair proteins through interaction with the front side of proliferating cell 37. Touitou, R. et al. A degradation signal located in the C-terminus of p21WAF1/ nuclear antigen. EMBO J. 17, 2412–2425 (1998). CIP1 is a binding site for the C8 alpha-subunit of the 20S proteasome. EMBO J. 7. Gulbis, J. M., Kelman, Z., Hurwitz, J., O’Donnell, M. & Kuriyan, J. Structure of 20, 2367–2375 (2001). the C-terminal region of p21(WAF1/CIP1) complexed with human PCNA. Cell 38. Churchill, M. E. & Suzuki, M. ’SPKK’ motifs prefer to bind to DNA at A/T-rich 87, 297–306 (1996). sites. EMBO J. 8, 4189–4195 (1989). 8. Hingorani, M. M. & O’Donnell, M. Sliding clamps: a (tail)ored fit. Curr. Biol. 39. Arya, G. & Schlick, T. A tale of tails: how histone tails mediate chromatin 10, R25–R29 (2000). compaction in different salt and linker histone environments. J. Phys. Chem. A 9. Yu, P. et al. p15(PAF), a novel PCNA associated factor with increased 113, 4045–4059 (2009). expression in tumor tissues. Oncogene 20, 484–489 (2001). 40. Kochaniak, A. B. et al. Proliferating cell nuclear antigen uses two distinct modes 10. Emanuele, M. J., Ciccia, A., Elia, A. E. & Elledge, S. J. Proliferating cell nuclear to move along DNA. J. Biol. Chem. 284, 17700–17710 (2009). antigen (PCNA)-associated KIAA0101/PAF15 protein is a cell cycle-regulated 41. Laurence, T. A. et al. Motion of a DNA sliding clamp observed by single anaphase-promoting complex/cyclosome substrate. Proc. Natl Acad. Sci. USA molecule fluorescence spectroscopy. J. Biol. Chem. 283, 22895–22906 (2008). 108, 9845–9850 (2011). 42. Randell, J. C. & Coen, D. M. Linear diffusion on DNA despite high-affinity 11. Povlsen, L. K. et al. Systems-wide analysis of ubiquitylation dynamics reveals a binding by a DNA polymerase processivity factor. Mol. Cell 8, 911–920 (2001). key role for PAF15 ubiquitylation in DNA-damage bypass. Nat. Cell Biol. 14, 43. Actis, M. et al. Small molecule inhibitors of PCNA/PIP-box interaction 1089–1098 (2012). suppress translesion DNA synthesis. Bioorg. Med. Chem. 21, 1972–1977 (2013). 12. Xie, C., Yao, M. & Dong, Q. Proliferating cell unclear antigen-associated factor 44. Punchihewa, C. et al. Identification of small molecule proliferating cell nuclear (PAF15): a novel oncogene. Int. J. Biochem. Cell Biol. 50C, 127–131 (2014). antigen (PCNA) inhibitor that disrupts interactions with PIP-box proteins and 13. Hosokawa, M. et al. Oncogenic role of KIAA0101 interacting with proliferating inhibits DNA replication. J. Biol. Chem. 287, 14289–14300 (2012). cell nuclear antigen in pancreatic cancer. Cancer Res. 67, 2568–2576 (2007). 45. De Biasio, A. et al. Reduced stability and increased dynamics in the human 14. Kais, Z. et al. KIAA0101 interacts with BRCA1 and regulates centrosome proliferating cell nuclear antigen (PCNA) relative to the yeast homolog. PLoS number. Mol. Cancer Res. 9, 1091–1099 (2011). ONE 6, e16600 (2011). 15. Kato, T. et al. Overexpression of KIAA0101 predicts poor prognosis in primary 46. Wishart, D. S. et al. 1H, 13C and 15N chemical shift referencing in biomolecular lung cancer patients. Lung Cancer 75, 110–118 (2012). NMR. J. Biomol. NMR 6, 135–140 (1995). 16. Liu, L. et al. Variant 1 of KIAA0101, overexpressed in hepatocellular 47. Palacios, A. et al. The dimeric structure and the bivalent recognition of carcinoma, prevents doxorubicin-induced apoptosis by inhibiting p53 H3K4me3 by the tumor suppressor ING4 suggests a mechanism for enhanced activation. Hepatology 56, 1760–1769 (2012).

NATURE COMMUNICATIONS | 6:6439 | DOI: 10.1038/ncomms7439 | www.nature.com/naturecommunications 11 & 2015 Macmillan Publishers Limited. All rights reserved. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7439

targeting of the HBO1 complex to chromatin. J. Mol. Biol. 396, 1117–1127 S. Pongor for help with fluorescence polarization, N. Mailand for pointing out the p15 (2010). sequence similarity with histone H3 and S. Onesti for helpful discussions. This work was 48. Palacios, A. et al. Molecular basis of histone H3K4me3 recognition by ING4. supported by the Spanish MINECO grants CTQ2011-28680 to F.J.B., BFU2011-23815 to J. Biol. Chem. 283, 15956–15964 (2008). G.M., and Juan de la Cierva to A.De.B., grant NNF14CC0001 to G.M., and SPIN-HD– 49. Kabsch, W. Xds. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132 (2010). ANR Chaires d’Excellence and the ATIP-Avenir to P.B. This research received funding 50. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, from the European Community’s FP7 grant agreement number 283570 (BioStruct-X, 658–674 (2007). providing acces to SLS and ALBA syncrotrons) and 261863 (Bio-NMR, providing access 51. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development to the 1 GHz NMR spectrometer at the Rhoˆne-Alpes European Large Scale Facility for of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010). NMR). 52. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010). Author contributions 53. Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from A.De.B. and F.J.B. conceived and designed the study. A.De.B., N.M. and M.V. purified crystalline state. J. Mol. Biol. 372, 774–797 (2007). the proteins. A.De.B., A.I.de.O., T.D. and F.J.B. performed and analysed the NMR 54. Bernado´,P.et al. A structural model for unfolded proteins from residual experiments. A.De.B. performed the fluorescence polarization experiments. G.B.M. dipolar couplings and small-angle x-ray scattering. Proc. Natl Acad. Sci. USA crystallized the complex, R.M. and G.M. solved and refined the crystal structure. F.C. 102, 17002–17007 (2005). performed the ITC experiments, F.C. and I.L. the ITC experiments. S.D. prepared 55. Eyal, E., Najmanovich, R., McConkey, B. J., Edelman, M. & Sobolev, V. samples for EM, took micrographs and selected individual particles. D.G.-C. classified Importance of solvent accessibility and contact surfaces in modeling side-chain and analysed the particles. T.N.C. and P.B. performed the molecular modelling. conformations in proteins. J. Comput. Chem. 25, 712–724 (2004). A.De.B. and F.J.B. wrote the paper. 56. van Dijk, M., van Dijk, A. D., Hsu, V., Boelens, R. & Bonvin, A. M. Information-driven protein-DNA docking using HADDOCK: it is a matter of flexibility. Nucleic Acids Res. 34, 3317–3325 (2006). Additional information 57. Flores-Rozas, H. et al. Cdk-interacting protein 1 directly binds with Accesion numbers: Coordinates and structure factors of the p1550–77–PCNA complex proliferating cell nuclear antigen and inhibits DNA replication catalyzed by the are deposited in the Protein Data Bank under accession code 4D2G. The assignment of DNA polymerase delta holoenzyme. Proc. Natl Acad. Sci. USA 91, 8655–8659 p15 bound to PCNA is deposited in the BioMagResBank with accession code 19868. (1994). Supplementary Information accompanies this paper at http://www.nature.com/ 58. Sorzano, C. O. et al. XMIPP: a new generation of an open-source image naturecommunications processing package for electron microscopy. J. Struct. Biol. 148, 194–204 (2004). Competing financial interests: The authors declare no competing financial interests. 59. Pascual-Montano, A. et al. A novel neural network technique for analysis and classification of EM single-particle images. J. Struct. Biol. 133, 233–245 (2001). Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions.

Acknowledgements How to cite this article: De Biasio, A. et al. Structure of p15PAF–PCNA complex and We thank P. Redondo for help with the DNA duplex formation, T. Kaminishi for help implications for clamp sliding during DNA replication and repair. Nat. Commun. 6:6439 with molecular modelling, M. Lelli for help with NMR relaxation measurements, doi: 10.1038/ncomms7439 (2015).